# Token Factory Docs > Token Factory is Omniva's managed inference service — managed inference, tuned for every token. In Early Access today: serverless, pay-per-token access to optimized open models through an OpenAI-compatible API at https://api.tokenfactory.omniva.com/v1. Dedicated Endpoints — reserved GPU capacity for eligible models, billed per GPU-hour — are also available. This file indexes the public docs for AI agents and crawlers. ## Getting Started - [Workspace access](https://tokenfactory.omniva.com/docs/getting-started/workspace-access): Sign in to your workspace, find your workspace ID, get help. - [Quickstart](https://tokenfactory.omniva.com/docs/getting-started/quickstart): Make your first OpenAI-compatible API call. - [Authentication](https://tokenfactory.omniva.com/docs/getting-started/authentication): Bearer tokens, key creation, key rotation. - [OpenAI-compatible migration](https://tokenfactory.omniva.com/docs/getting-started/openai-compatible-migration): Swap base URL + key, what works unchanged, what differs. - [Models overview](https://tokenfactory.omniva.com/docs/getting-started/models-overview): What models we offer, sort and filter conventions. ## Serverless - [Chat completions](https://tokenfactory.omniva.com/docs/serverless/chat-completions): Concept, minimal example, streaming, tools. - [Embeddings](https://tokenfactory.omniva.com/docs/serverless/embeddings): Vector representations for retrieval and similarity. - [Streaming](https://tokenfactory.omniva.com/docs/serverless/streaming): Server-Sent Events format and language parsers. - [Tokens, pricing & quotas](https://tokenfactory.omniva.com/docs/serverless/pricing): How tokens are counted, billed, and capped. ## Dedicated Endpoints - [What are Dedicated Endpoints?](https://tokenfactory.omniva.com/docs/dedicated-endpoints/what-are-dedicated-endpoints): Reserved GPU capacity vs. shared serverless — what changes, what stays the same. - [Create a dedicated endpoint](https://tokenfactory.omniva.com/docs/dedicated-endpoints/create-endpoint): Name, model, flavor, and scaling — start to finish. - [Call a dedicated endpoint](https://tokenfactory.omniva.com/docs/dedicated-endpoints/call-endpoint): Route requests with the dedicated/ Model ID. - [Autoscaling](https://tokenfactory.omniva.com/docs/dedicated-endpoints/autoscaling): The replica envelope, target metric, and scale windows. - [GPU allotment](https://tokenfactory.omniva.com/docs/dedicated-endpoints/gpu-allotment): How per-org GPU capacity is capped, read, and what a 409 means. - [Lifecycle](https://tokenfactory.omniva.com/docs/dedicated-endpoints/lifecycle): Suspend, resume, edit scaling, and delete — and what each does to your allotment. - [Monitor dedicated endpoints](https://tokenfactory.omniva.com/docs/dedicated-endpoints/monitoring): Fleet health, per-endpoint drill-down, and GPU cost. - [GPU cost](https://tokenfactory.omniva.com/docs/dedicated-endpoints/cost): Billable GPU spend, hours, and utilization across your dedicated fleet. ## API Reference - [Chat completions](https://tokenfactory.omniva.com/docs/api-reference/chat-completions): Create a chat completion. - [Embeddings](https://tokenfactory.omniva.com/docs/api-reference/embeddings): Create embeddings for one or more inputs. - [Models](https://tokenfactory.omniva.com/docs/api-reference/models): List available models. - [Endpoints](https://tokenfactory.omniva.com/docs/api-reference/endpoints): Create, list, retrieve, update, suspend, resume, and delete dedicated endpoints. - [GPU allotment headroom](https://tokenfactory.omniva.com/docs/api-reference/gpu-allotment-headroom): Read your organization's GPU allotment usage and remaining headroom. ## Reference - [Errors & status codes](https://tokenfactory.omniva.com/docs/reference/errors): Every error code with cause and fix, including SSL chain. - [Observability metrics](https://tokenfactory.omniva.com/docs/reference/observability-metrics): Every metric defined in plain English. - [Glossary](https://tokenfactory.omniva.com/docs/reference/glossary): Terms and concepts used across these docs, defined once. ## Conventions - API base URL: `https://api.tokenfactory.omniva.com/v1`. - Authentication: bearer token in the `Authorization` header. See the Authentication page. - Compatibility: the inference surface targets the OpenAI v1 contract. Differences are documented per page. ## Retrieval - Any page above is available as raw Markdown by appending `.md` to its URL — e.g. https://tokenfactory.omniva.com/docs/getting-started/quickstart.md - The same source is served at `/api/docs/raw/
/`, which the `.md` form rewrites to. - The entire documentation set in one response: https://tokenfactory.omniva.com/llms-full.txt - This index is also served at https://tokenfactory.omniva.com/llms.txt